Chapter 5: Data Mining for Business Intelligence

Question 1

Apriori algorithm

Answer 1

The most frequently used algorithm to find association rules. This algorithm identifies subsets that are frequent to a minimum number of the itemsets. The frequent subsets are extended one item at a time. This means it will increase from one-item subsets to two-item subsets, then three-item subsets and so on until there are no more successful extensions found.

Question 2

Area under the ROC curve

Answer 2

It is a graphical plot in which the true positive rate is plotted on the Y- axis and the false positive rate is plotted on the X-axis. The area under the ROC curve deter determines the accuracy measure of a classifier: value1 indicates a perfect classifier, a value of 0.5 indicates no better than a random chance.

Question 3

Associations (Association rule learning in data mining)

Answer 3

It is a research technique that is used to identify relationships among variables in the database. In the retail industry, associations rule mining is called market basket analysis. Link analysis and sequence mining are derivatives of association rule mining.

Question 4

Bootstrapping

Answer 4

is when a fixed number of instances from the original data is sampled (with replacement) for training and the rest of the data set is used for testing. This process can be repeated if need it.

Question 5

Categorical data

Answer 5

are labels for classes that are used to divide a variable into specific groups. Categorical data is also call discrete data mining that is represents finite number of values with no continuum between them.

Question 6

Classification

Answer 6

or supervised induction. It is a very common data mining task. This task analyzes the historical data and generates a model that can predict future behavior. This model consists of generalizations over the records of a training dataset, which help distinguish predefined classes. The expectation is that this model can be used to predict the classes of other unclassified records and even to predict actual future events.

Question 7

Clustering

Answer 7

is the process of partitioning a collection of objects, events, etc. presented in a dataset, into natural groups (sub-classes) where the members share similar characteristics. Commonly used clustering techniques are k-means (in statistics) and self-organizing maps (in machine learning).

Question 8

Confidence

Answer 8

is one of the metrics that association rules mining uses to answer the question: “Are all association rules interesting and useful?” Confidence measures how often consequent go together with antecedent.

Question 9

CRISP-DM

Answer 9

(Cross Industry Standard Process for Data Mining) is a general process for doing data mining projects.

Question 10

Data mining

Answer 10

in the discovering of patterns and significant knowledge in large quantity of data.

Question 11

Decision tree

Answer 11

builds classification or regression models in the form of a tree structure. It breaks down data into smaller and smaller subsets. Input variables in a decision tree are called attributes. A tree has branches and nodes; a branch represents the outcome of a test to classify a pattern, and each leaf node holds a class label. The topmost node in the tree is the root node.

Question 12

Discovering-driven data mining

Answer 12

is a technique used to find patterns, associations, and other relationships hidden within datasets. It usually discovers facts that organization had not previously known.

Question 13

Distance measure

Answer 13

is used in cluster analysis methods to calculate the closeness between pairs of items. Well known distance measures are Euclidian distance (distance between two points that can be measured with a ruler) and the Manhattan distance (the rectilinear distance, or taxicab distance, between two points).

Question 14

Entropy

Answer 14

measures the extent of uncertainty or randomness in a dataset.

Question 15

Gini index

Answer 15

can be used to determine the purity of a specific class as a result of a decision to branch along a particular attribute or variable.

Question 16

Hypothesis-driven data mining

Answer 16

it is a technique that begins with a proposition by the user, who then seeks to validate the truthfulness of the proposition.

Question 17

Information gain

Answer 17

is the splitting mechanism used in ID3.The most well-known decision tree algorithm.

Question 18

Interval data

Answer 18

are variables that can be measured on interval scales. For example: temperature on the Celsius scale.

Question 19

k-fold cross-validation

Answer 19

or rotation estimation is when the original sample is randomly partitioned into k equal size subsets. Then the classification model is train and tested k times. Each time is trained in all but one-fold and then tested on the remaining single fold.

Question 20

Knowledge discovery in databases (KDD)

Answer 20

has been defined as a process of using data mining methods (involving the use of algorithms to identify patterns in data derived through the KDD process. The KDD process consist on: data selection, data processing, data transformation, data mining, and interpretation (evaluation).

Question 21

Link analysis

Answer 21

when doing link analysis, the linkage among many objects is discovered automatically.

Question 22

Microsoft Enterprise Consortium

Answer 22

is the worldwide source for access to Microsoft’s SQL Server 2008 software suite for academic purposes (teaching and research). The consortium provides wide range of business intelligence developing tools and different large datasets.

Question 23

Microsoft’s SQL Server

Answer 23

is a suite of business intelligence capabilities that has become very popular for data mining because data and the models are stored in the same relational database environment. This makes model management an easier task.

Question 24

Nominal data

Answer 24

contains code assigned to objects as labels, which are not measurements. Example for the variable marital status the general categories are: single, married, and divorced. Nominal data can be represented by binominal values like yes/no, true/false, good/bad. Nominal data can be represented by multinomial values like: brown/green/blue, another example is: single/married/divorced.

Question 25

Numeric data

Answer 25

is the numerical value of the variable. Numerical variables are age, number of children, household income (in dollars), travel distance (in miles), temperature (in Fahrenheit degrees). Numeric data is also called continuous data because represents scalable measurements.

Question 26

Ordinal data

Answer 26

contains code assigned to objects or events as labels, this labels also represent the rank order among them. Example credit score can be: low, medium, or high.

Question 27

Prediction

Answer 27

is used as a synonym of forecasting in data mining. Prediction can be named as classification where the predicted object, tomorrow forecast is a class label as “rainy” or “sunny”. The prediction could be a regression where the predicted thing is tomorrow’s temperature, then is a real number like “70 °F.”

Question 28

RapidMiner

Answer 28

Is a free data mining tool. Its graphical user interface, employment of large number of algorithms, and the incorporation of different data visualization features is making it stand out from other fee tools.

Question 29

Ratio data

Answer 29

are measurement variables found in the physical sciences and engineering. For example, mass, length, time, plane angle, energy, electric charge.

Question 30

Regression

Answer 30

is a data mining method in what is being predicted is a numeric value like a temperature (“70 °F “).

Question 31

SAS Enterprise Miner

Answer 31

is a free data mining tool which has growth in popularity among commercial tools.

Question 32

SEMMA

Answer 32

(Sample, Explore, Modify, Model, and Assess). The SEMMA data mining process starts with the generation of a representative sample of the data, which makes easier to apply exploratory statistical and visualization techniques, then select and transform (modify) the most significant predictive variables, then model the variables to predict outcomes, and finally conduct assessments to evaluate the accuracy and usefulness of the models.

Question 33

Sequence mining

Answer 33

is the other derivative of association rule mining. In sequence mining relationships are examined in terms of their order of occurrence to identify associations over time.

Question 34

Simple split

Answer 34

is a hold-out method for testing model accuracy. The simple split partitions the data in two exclusive subsets called training set and test set (or holdout set).

Question 35

SPSS PASW Modeler

Answer 35

is the most popular data mining software tool according to the May 2009 survey by Kdnuggets.com

Question 36

Support

Answer 36

is a metric for association rule that measures how often the antecedents and the consequents appear together in the same transaction.

Question 37

Weka

Answer 37

is the most popular free and open source software tool for data mining. It was developed by researchers from the University Waikato in New Zealand.

Question 38

Reasons for businesses growing interest in data mining

Answer 38

o Intense competition driven by customer constantly changing needs
o Recognition of the hidden values in large data sources.
o The consolidation and integration of database records.
o Being able to place databases and data repositories in a data warehouse.
o The rapid increase on speed of processing and storage.
o The cost of hardware and software for storing data and processing it have been decrease.

Question 39

Prominent characteristics and objectives of data mining

Answer 39

 There are data which needs to be sorted from large databases. It could contain years of accumulated data. In most cases the data has been cleaned and synthesize into a data warehouse.
 Client-server environment or Web-based Information Systems architecture.
 New tools that facilitates interaction with user
 The end user is the miner in most cases
 The miner needs to have a creative thinking and the capability to interpret the findings.
 Other characteristic of data mining is that its tools are already working with spreadsheets and software development products.
 Another characteristic is the need to use parallel processing to handle the enormous amount of data and the search.

Question 40

Data Classification in Data Mining

Answer 40

• Categorical Data
• Nominal Data
• Ordinal Data
• Numerical Data
• Interval Data
• Ratio Data

Question 41

data mining techniques

Answer 41

o Prediction
o Association
o Clustering

Question 42

Data Mining Applications

Answer 42

• Since data mining has been used to address many complicated businesses problems it has increase in popularity. Business is utilizing data mining to resolve serious problems affecting them currently and other times to explore opportunities of gaining competitive advantage. Here are some ways different businesses are using data mining:
  o Customer relationship management: for identifying most profitable customers for example
  o Banking: automating the loan application process, and predicting the most probable defaulters.
  o Retailing and logistics: predicting sales in order to determine inventory
  o Manufacturing and production: predicting machine failure
  o Brokerage and securities trading: forecasting the range and direction of stock fluctuations
  o Insurance: forecast claim amounts for property and medical coverage costs
  o Computer hardware and software: to predict disk drive failures
  o Government and defense: forecasting the cost of moving military personnel and equipment and predict adversary moves.
  o Travel industry (airlines, hotels/resorts, rental car companies): to predict sales of different services.
  o Health care: identifying people without insurance and the factors for it.
  o Medicine: predict success rates of organ transplants

Question 43

Data Mining Process

Answer 43

• Cross-Industry Standard Process for Data Mining (CRISP-DM) proposed that there are six steps in data mining process.
  o Business understanding: is to know what the study id for
  o Data understanding: is the identification of the relevant data
  o Data preparation: also called data processing is to take data identified in previous step and prepare it for analysis by data mining methods (cleaning data etc.)
  o Model building: assessment and comparative analysis of various models built. Modeling techniques are selected and applied
  o Testing and evaluation: developed models are assessed and evaluated for their accuracy
  o Deployment: this is the last step and is here where the knowledge gained from the data mining process is presented in a way the user can understand and benefit from these findings
• Other data mining standardized process is SEMMA. SEMMA (Sample, Explore, Modify, Model, and Assess). The SEMMA data mining process starts with the generation of a representative sample of the data, which makes easier to apply exploratory statistical and visualization techniques, then select and transform (modify) the most significant predictive variables, then model the variables to predict outcomes, and finally conduct assessments to evaluate the accuracy and usefulness of the models.

Question 44

Classification Techniques

Answer 44

o	Decision tree analysis
o	statistical analysis
o	Neural networks
o	Case-based reasoning
o	Bayesian classifiers
o	Genetic algorithms
o	Rough sets